Skip to contentSkip to conversation

libxl: slightly correct JSON generation of CPU policy

PATCH In review Jan Beulich 3 participants 5 messages Aug 15, 2023

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 libxl: slightly correct JSON generation of CPU policy

The "cpuid_empty" label is also (in principle; maybe only for rubbish input) reachable in the "cpuid_only" case. Hence the label needs to live ahead of the check of the variable.

Fixes 5b80cecb747b ("libxl: introduce MSR data in libxl_cpuid_policy")
tools/libs/light/libxl_cpuid.c +2 -1
@@ -710,10 +710,11 @@ parse_cpuid:
                     libxl__strdup(NOGC, libxl__json_object_get_string(r));
         }
     }
+
+cpuid_empty:
     if (cpuid_only)
         return 0;
 
-cpuid_empty:
     co = libxl__json_map_get("msr", o, JSON_ARRAY);
     if (!libxl__json_object_is_array(co))
         return ERROR_FAIL;
Reviewed-by on v1 1/1
#1 lore raw
Quoted 6 lines · Jan Beulich

The "cpuid_empty" label is also (in principle; maybe only for rubbish input) reachable in the "cpuid_only" case. Hence the label needs to live ahead of the check of the variable.

Fixes 5b80cecb747b ("libxl: introduce MSR data in libxl_cpuid_policy")

Thanks,

Signature · 1 line
Anthony PERARD
#2 lore raw
Quoted 8 lines · Anthony PERARD
Quoted 6 lines · Jan Beulich

The "cpuid_empty" label is also (in principle; maybe only for rubbish input) reachable in the "cpuid_only" case. Hence the label needs to live ahead of the check of the variable.

Fixes 5b80cecb747b ("libxl: introduce MSR data in libxl_cpuid_policy")

Hmm - this was included in the security backports.  I guess it ought to go back to 4.15 (now that 4.14 is properly dead).

~Andrew

#3 lore raw
Quoted 6 lines · Andrew Cooper
Quoted 5 lines · Anthony PERARD
Quoted 4 lines · Jan Beulich

The "cpuid_empty" label is also (in principle; maybe only for rubbish input) reachable in the "cpuid_only" case. Hence the label needs to live ahead of the check of the variable.

Quoted 6 lines
Quoted 3 lines
Quoted 2 lines
Fixes 5b80cecb747b ("libxl: introduce MSR data in libxl_cpuid_policy")

Hmm - this was included in the security backports.  I guess it ought to go back to 4.15 (now that 4.14 is properly dead).

Depends on whether this is just a theoretical issue, I would say. Can legacy format input legitimately be empty? If not, I wouldn't bother putting this onto the security-only branches. (Guess how I found it. For the 4.13 backport much of this needed quite a bit of adjustment. Pulling in more stuff from 4.14 isn't really an option, not the least since I will need to [at least try to] go further back.)

Jan

#4 lore raw
Quoted 9 lines · Jan Beulich
Quoted 8 lines · Andrew Cooper
Quoted 7 lines · Anthony PERARD
Quoted 6 lines · Jan Beulich

The "cpuid_empty" label is also (in principle; maybe only for rubbish input) reachable in the "cpuid_only" case. Hence the label needs to live ahead of the check of the variable.

Fixes 5b80cecb747b ("libxl: introduce MSR data in libxl_cpuid_policy")
Quoted 8 lines
Quoted 4 lines
Quoted 1 line

Hmm - this was included in the security backports.  I guess it ought to go back to 4.15 (now that 4.14 is properly dead).

Depends on whether this is just a theoretical issue, I would say. Can legacy format input legitimately be empty? If not, I wouldn't bother putting this onto the security-only branches.

I don't think libxl would generate an empty array. libxl calls libxl__cpuid_policy_is_empty() before calling libxl_cpuid_policy_list_gen_json(), so no empty array.

Then, without this fix, I think it would just mean that libxl would return an error on something that was ok to do before. So backporting the patch would be a nice to have.

Signature · 1 line
Anthony PERARD